Evaluation of baseflow separation methods with real and synthetic streamflow data from a watershed

Abstract

• Performance of graphical & digital filter baseflow separation methods are compared. • Methods are compared using real data with a subjective concept of hydrologic plausibility. • Synthetic data from an integrated hydrologic model used for more rigorous assessments. • Estimated baseflow hydrographs reveal seasonal variability. • Filtered UKIH approach works best for most study locations but depends spatially. Baseflow originating primarily from groundwater is a critical streamflow component, although its accurate estimation is fraught with significant difficulties. This study estimates baseflow through existing graphical and digital filter methods, using actual streamflow data from a gauging station at the Alder Creek Watershed (ACW) and synthetic streamflow data at ten study locations within the same watershed simulated with HydroGeoSphere (HGS) (Aquanty Inc., 2018). There are four widely used graphical (Institute for Hydrology, 1980; Sloto and Crouse, 1996; Aksoy et al., 2008) and six digital filtering (Lyne and Hollick, 1979; Chapman and Maxwell, 1996; Furey and Gupta, 2001; Eckhardt, 2005; Tularam and Ilahee, 2008; Aksoy et al., 2009) baseflow separation approaches compared in this study. To determine the most optimal approach, baseflow estimates from real data are assessed based on the subjective concept of hydrologic plausibility, while baseflow estimates obtained from a HGS streamflow record with graphical and digital filtering methods are compared to those computed directly by HGS. Overall, results from this study indicate that baseflow hydrographs reveal a seasonal pattern at the ACW. During wintertime, streamflow is composed almost entirely of baseflow, whereas during summertime, baseflow only consists approximately 20% to 60% of streamflow. After comparing baseflow estimates with those computed by HGS, the most optimal approaches at the ten study locations are assessed. Results show that the best approach at six study locations is the FUKIH (Aksoy et al., 2009) approach, while at three locations, the Chapman and Maxwell (1996) approach and for one location, the Eckhardt (2005) approach performed the best. In conclusion, it is inferred that the most optimal approach within the ACW varies spatially.